Running multiple agents in parallel is bringing back context-switching fatigue, wondering if anyone found a real fix?

Not asking for prompt/agent.md tips, I'm past that — I run an entire department where I'm supposed to guide others, so this is more about workflow at scale than tuning individual agent behavior. 16+ years into this career, and writing code isn't the hard part anymore — let alone with the codebase changes AI makes on its own. Before, every decision used to go on a whiteboard and get documented, so if I wrote it once, I'd never forget it. What's hard now is running multiple Claude/agent instances across different projects at once and still feeling like I'm the bottleneck. My old manager used to call me "devil in the details" — I pay attention to everything, which used to be an asset. Now it means I feel every single context switch. Went down the rabbit hole: swapped terminal multiplexers, started building a kanban ( because why not ) just for agent task tracking — stalled out because I realized I was missing something more fundamental, not a UI problem. Tried Herdr as a tmux-for-agents thing to be on the same hype wave. Feels like everyone's solving the tooling layer and not the "how do I actually manage my own attention across N parallel AI workstreams" layer. How's everyone actually living with this — not the demo version, the real day-to-day? Thanks in advance!